Open Access

Genome-wide association studies (GWAS) are used to identify genetic markers linked with at least partially heritable diseases or phenotypes without prior knowledge of any disease-associated genetic loci. In summer 2008, all individuals of the population based cohort Study of Health in Pomerania (SHIP) were individually genotyped using the Affymetrix Genome-Wide Human SNP Array 6.0 microarray. The aim of this work was to establish an efficient workflow for GWAS using the more than 4000 individually genotyped samples of the SHIP cohort as well as pooled samples, focusing exclusively on analyzing genetic variations based on single nucleotide polymorphisms (SNPs). Firstly, an optimal array platform for the genotyping analysis had to be chosen that detected most of the available genetic variants at a high level of accuracy. Secondly, extensive quality controls had to be performed starting from DNA extraction and including tests of the generated array data by the analysis software to obtain the most reliable data for the subsequent association studies. For the identification of loci with smaller genetic influences, individual cohorts were meta-analyzed in large nationally and internationally organized consortia (e.g. CHARGE, BPGen, HaemGen, GIANT, CKD Gen). To participate in those meta-analyses, a comparable common set of genetic data had to be generated. This was done by imputation of the data generated by individual array-based genotyping on the basis of a reference panel using chromosomal linkage information. Due to the extensive phenotype information in the SHIP study, it was possible to perform many genome-wide discovery analyses and replication studies of possible susceptibility loci in a short time once the genetic data was available and processed. This resulted in the necessity to set up an efficient workflow for storing the huge amount of genetic data, converting it into different formats readable for specific analysis software, performing the association analyses and processing the results into a human-readable and clear format. This included replications, GWAS and meta-analyses of several cohorts. Many susceptibility loci were newly identified in different association studies with the SHIP data included and were subsequently published. In this work, genetic association studies with the SHIP data included were performed and published on blood pressure, uric acid concentrations, cardiac structure and function, lipid metabolism, hematological parameters, kidney functions, smoking quantity, circulating IGF-I and IGFBP-3 concentrations and thyroid volume including the risk of goiter development. Besides the SHIP cohort, there was a need to use other, especially patient cohorts for GWAS. Since no genotype information from these patient cohorts was available and the individual genotyping of many probands is still expensive and therefore often not affordable, we established the cost-effective allelotyping method that relied on pooling of DNA samples prior to the hybridization with microarrays. After estimating the pooling-specific error of a case-control allelotyping study, the allelotyping approach was used for identifying genetic susceptibility loci associated with aggressive periodontitis. If not referring to work of collaborators, all statistical analyses, data handling and in silico work concerning the SHIP data described in this context was performed by the author of this dissertation.

Activation of trace amine-associated receptor 1 (TAAR1) in endocrine pancreas is involved in weight regulation and glucose homeostasis. The purpose of this study was the identification and characterization of potential TAAR1 variants in patients with overweight/obesity and disturbed glucose homeostasis. Screening for TAAR1 variants was performed in 314 obese or overweight patients with impaired insulin secretion. The detected variants were functionally characterized concerning TAAR1 cell surface expression and signaling properties and their allele frequencies were determined in the population-based Study of Health in Pomerania (SHIP). Three heterozygous carriers of the single nucleotide missense variants p.Arg23Cys (R23C, rs8192618), p.Ser49Leu (S49L, rs140960896), and p.Ille171Leu (I171L, rs200795344) were detected in the patient cohort. While p.Ser49Leu and p.Ille171Leu were found in obese/overweight patients with slightly impaired glucose homeostasis, p.Arg23Cys was identified in a patient with a complete loss of insulin production. Functional in vitro characterization revealed a like wild-type function for I171L, partial loss of function for S49L and a complete loss of function for R23C. The frequency of the R23C variant in 2018 non-diabetic control individuals aged 60 years and older in the general population-based SHIP cohort was lower than in the analyzed patient sample. Both variants are rare in the general population indicating a recent origin in the general gene pool and/or the consequence of pronounced purifying selection, in line with the obvious detrimental effect of the mutations. In conclusion, our study provides hints for the existence of naturally occurring TAAR1 variants with potential relevance for weight regulation and glucose homeostasis.

Mendelian randomization (MR) is a framework for assessing causal inference using cross-sectional data in combination with genetic information. This paper summarizes statistical methods commonly applied and strait forward to use for conducting MR analyses including those taking advantage of the rich dataset of SNP-trait associations that were revealed in the last decade through large-scale genome-wide association studies. Using these data, powerful MR studies are possible. However, the causal estimate may be biased in case the assumptions of MR are violated. The source and the type of this bias are described while providing a summary of the mathematical formulas that should help estimating the magnitude and direction of the potential bias depending on the specific research setting. Finally, methods for relaxing the assumptions and for conducting sensitivity analyses are discussed. Future researches in the field of MR include the assessment of non-linear causal effects, and automatic detection of invalid instruments.

A decreased estimated glomerular filtration rate (eGFR) leading to chronic kidney disease is a significant public health problem. Kidney function is a heritable trait, and recent application of genome-wide association studies (GWAS) successfully identified multiple eGFR-associated genetic loci. To increase statistical power for detecting independent associations in GWAS loci, we improved our recently developed quasi-adaptive method estimating SNP-specific alpha levels for the conditional analysis, and applied it to the GWAS meta-analysis results of eGFR among 783,978 European-ancestry individuals. Among known eGFR loci, we revealed 19 new independent association signals that were subsequently replicated in the United Kingdom Biobank (n = 408,608). These associations have remained undetected by conditional analysis using the established conservative genome-wide significance level of 5 × 10–8. Functional characterization of known index SNPs and novel independent signals using colocalization of conditional eGFR association results and gene expression in cis across 51 human tissues identified two potentially causal genes across kidney tissues: TSPAN33 and TFDP2, and three candidate genes across other tissues: SLC22A2, LRP2, and CDKN1C. These colocalizations were not identified in the original GWAS. By applying our improved quasi-adaptive method, we successfully identified additional genetic variants associated with eGFR. Considering these signals in colocalization analyses can increase the precision of revealing potentially functional genes of GWAS loci.

Background Approaching epidemiological data with flexible machine learning algorithms is of great value for understanding disease-specific association patterns. However, it can be difficult to correctly extract and understand those patterns due to the lack of model interpretability. Method We here propose a machine learning workflow that combines random forests with Bayesian network surrogate models to allow for a deeper level of interpretation of complex association patterns. We first evaluate the proposed workflow on synthetic data. We then apply it to data from the large population-based Study of Health in Pomerania (SHIP). Based on this combination, we discover and interpret broad patterns of individual serum TSH concentrations, an important marker of thyroid functionality. Results Evaluations using simulated data show that feature associations can be correctly recovered by combining random forests and Bayesian networks. The presented model achieves predictive accuracy that is similar to state-of-the-art models (root mean square error of 0.66, mean absolute error of 0.55, coefficient of determination of R2 = 0.15). We identify 62 relevant features from the final random forest model, ranging from general health variables over dietary and genetic factors to physiological, hematological and hemostasis parameters. The Bayesian network model is used to put these features into context and make the black-box random forest model more understandable. Conclusion We demonstrate that the combination of random forest and Bayesian network analysis is helpful to reveal and interpret broad association patterns of individual TSH concentrations. The discovered patterns are in line with state-of-the-art literature. They may be useful for future thyroid research and improved dosing of therapeutics.